Open stefnotch opened 1 year ago
Hey, yes I would definitely consider it. Some care would need to be taken though or it might be a bit of a waste... I think ideally, when the rkyv feature flag is enabled all the types would have their <T as rkyv::Archive>::Archived
type be Self
. That way you could do math operations directly on the loaded archived form. Otherwise if you have to deserialize to actually do anything it would be a bit of a waste of rkyv
's power.
To accomplish that properly we'd need to also add #[repr(C)]
on things when the rkyv feature is on and maybe even add support to the underlying vector types in wide
, if we want to support simd wide vectors too.
Hey, yes I would definitely consider it. Some care would need to be taken though or it might be a bit of a waste... I think ideally, when the rkyv feature flag is enabled all the types would have their
<T as rkyv::Archive>::Archived
type beSelf
. That way you could do math operations directly on the loaded archived form.
That would definitely be awesome. I'm not certain if that's fully possible at the moment, due to endian-ness https://github.com/rkyv/rkyv/issues/292 .
(And sadly the official documentation chapter 18.1. Impl duplication
hasn't been written yet either https://rkyv.org/ )
I was wondering if you'd consider adding rkyv support to this library, of course behind a feature flag?
Hopefully all that would need to be done is deriving the types, and everything works. I would also prepare a PR with that if wanted.